Texture descriptors and machine learning algorithms for mistletoe detection in urban forests using multispectral imagery

Abstract

In this study, we compare the performance of texture descriptors and spectral vegetation indices for the classification of a hemiparasitic plant that grows on host trees, known as mistletoe. For this purpose, we computed 180 image features, including GLCM, Gabor, and LBPs, as well as spectral vegetation indices, from multispectral aerial image sets. Our image feature database is then classified using Support Vector Machines, with optimized hyperparameters, and accuracy metrics are reported in order to evaluate the contribution of specific feature sets for our application. In addition, we make use of feature selection algorithms in order to determine which combination of descriptors improves the classification process. The study has important implications for the remote sensing community, as it can provide insights into the use of texture and spectral descriptors for classification of the mistletoe species known as Struthanthus Interruptus. The results of the study can be used to develop more effective tools to monitor the spread of the pest in urban parks, which can help to preserve trees and ensure their long-term health. Overall, the study contributes to the growing body of research on the use of remote sensing technologies, in conjunction with artificial intelligence techniques, to monitor urban environments.